I used the SDF Reader to input an SDFile with Structure and 2 properties. I then immediately wrote it back out with SDF Write and only asked for the Structure column (I left the two input property columns as not selected). The output SDFile however had all the input columns (Structure and 2 Properties). Remarkably this is true even if I put a column filter in between the Input and Output nodes completely removing the property columns from the pipeline. It is as if the SDFile writer knows about all the input property columns and outputs them regardless, even though they are not in the pipeline anymore. [Knime V 2.7.1]
[ I discovered this because in the real work flow I had renamed one of the properties with Column Rename. I then matched with a right outer join to another input file and row split to get match and no match. When I then output an SDFile asking for the renamed column I also got a duplicate column with the original column name. A little detective work led to the much simpler statement of the issue given above. ]
Mark